Re: Hub Operations

["Michael McSharry" <mcs101main@xxxxxxx>]

------=_NextPart_000_0018_01C4B2B8.F3016CE0
Content-Type: text/plain;
charset="iso-8859-1"
Content-Transfer-Encoding: quoted-printable

MessagePort 3639 is someting that needs to be managed and if an
application=
claims responsibility for it, then I thing it also needs to take-on the ro=
le of hub function.  One that takes the candy and does not share is not
wel=
l-liked in the playground.=20=20

In my xap work I'm trying to manage redundancy to achieve higher
availabili=
y rates.  For me I want to have a hub which has only one role so if I take
=
action to manage that hub I will not need to also be concerned about other
=
functions that the hub may be performing.  If my hub happens to double as
a=
n HTTP gateway, then if I swap-out the hub I will also need to figure out
h=
ow to restore the HTTP gateway function.  While it all can be done, it
just=
adds to the complexity.  The nice thing about xap is how the functionality=
can be split into very small building blocks.  When these blocks start to =
do multiple things then the elegance is compromised.

----- Original Message -----=20
From: Patrick Lidstone (Personal e-mail)=20
To: xAP_developer@xxxxxxx=20
Sent: Friday, October 15, 2004 12:20 AM
Subject: RE: [xAP_developer] Hub Operations



I've been using the xapFramework hub and viewer and have the problems o=
f message duplication.  To better understand what is happening I wrote a
hu=
b application based upon what I understood from the spec and Patrick
Lidsto=
ne feedback.  With my hub I no longer get the duplicate messages.  It is
my=
understanding that the hub should use 127.0.0.1 to retransmit messages to =
all xap applicaitons on the same PC.  I see this as the From interface in
t=
he xapFramework viewer when I run my hub, but when I run the xapFramework
h=
ub I see 192.x.x.x. as the message From.  It looks as if the xapFramework
h=
ub is using the primary interface for retransmission.  I do not know which
=
is right or wrong, but I'm just trying to get my system working without
dup=
licate messages.  Is there concensus on how the hub should operate?

What is also interesting is that the viewer on machine X reports that i=
s sees the viewer on machine Y via  the 127.0.0.1:3639 interface.  I was
no=
t aware message sent on the loopback interface were visible on other
comput=
ers.  What does the xapFramework viewer use to determine the From column
in=
the display?=20

A message sent on the loopback interface categorically cannot be receiv=
ed by any other host under any circumstances.=20

In my hub implementation  messages received on the primary interface (p=
ort 3639) are sent on 127.0.0.1 to the subscribed port of each applicaton
o=
n the hub's PC using the port that the application indicated to which is
wa=
s bound in the heartbeat message.  All applications send their messages on
=
the primary interface to port 3639.  I think these messages will be seen
by=
the local hub and any other computers on the primary LAN.=20

This is the correct implementation.

Related question ....

In a PC desktop application is generally the case that more than one xa=
p application will be running.  If this is the case then why will the
deskt=
op application grap port 3639 and prevent normal xap opeations until
manual=
ly stopped?  It seems that a PC application should pick some other port
the=
n eventually become active when the hub function comes online.=20

Not sure if I have fully understood what you are saying here, but a hub=
less application running on a PC has no choice but to bind to 3639,
otherwi=
se it will be both deaf to incoming xAP messages, and no outgoing messages
=
be heard by other xAP applications because they will not be listening on
th=
e arbitrary port chosen by the app. There is a case to be made for
insistin=
g that all apps include built in hub functionality, although this
obviously=
raises the bar for developers. I think the majority of Stuart's apps take =
this approach; I haven't got round to it for the majority of my apps,
altho=
ugh I have the code in place, and will probably do so one day.

Patrick